Privacy is architecture, not policy.
Dataken captures desktop activity to power OLi. That is a sentence that requires a precise and defensible privacy story. Here it is.
Activity-graph anonymization
Dataken anonymizes the activity graph as an architectural property of the data layer. This is not a policy overlay or a configuration toggle — it is how the system is built. The activity graph stores structured work patterns, not personally identifiable browsing history.
Privatized LLM inference
Any skill or Ask OLi call that invokes an LLM uses the model provider's privatized-inference feature by default — zero retention, no training on tenant data. For extremely security-sensitive tenants, an open-source isolated-deployment option is available for organizations that commit to the infrastructure cost.
Per-tenant data boundaries
Each customer's rules registry, skills, and activity graph are isolated to that tenant. Cross-tenant data sharing is not a feature. Your data stays in your boundary.
On-device capture
The desktop agent captures and processes activity data on the user's machine. Only structured activity records — timestamp, application, dwell, content type — leave the device. No raw screenshots. No keylogging. No full-screen capture.
Consent your compliance team can defend.
OLi’s capture runs with user awareness and organizational consent. The system is designed so your DPO and compliance team can explain exactly what is collected, where it goes, and how to turn it off — because those answers are architectural, not contractual.
Talk to our security team →10% more productive time in 30 days. Or you don’t pay.
Book a 30-minute demoPrivacy questions
Frequently asked
- What exactly does Dataken capture from an employee's desktop?
- Application and task context — the type of work happening, its sequence, and its duration — anonymized at the point of capture into the activity graph. The system is designed to describe work patterns, not to reproduce the content of documents, messages, or screens.
- Where does the data go, and does it train external AI models?
- Activity data stays inside a per-tenant boundary, and model inference is privatized by default rather than routed to an external model provider. Customer activity data is not used to train third-party models.
- Will our security and compliance teams approve this?
- That review is the expected gate, not an obstacle to work around. The architecture — anonymization at capture, per-tenant data boundaries, privatized inference, and documented consent — is built specifically to pass second-line risk review in regulated industries including banking, insurance, healthcare, and defense.
- Do employees know OLi is running?
- Yes. Consent is captured at install, and OLi is visible to the user by design — it surfaces help at the point of work, so its presence is the product rather than something concealed. Deployments that hide the agent from employees are not something Dataken supports.